Re: xl allows multiple PCI attach of same device + triggers pciback bug(Was: Re: VMX status report. Xen:25605 & Dom0:3.4.4)

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Mon, Jul 23, 2012 at 11:38:02AM +0100, Ian Campbell wrote:
> On Wed, 2012-07-18 at 08:12 +0100, Wu, GabrielX wrote:
> > Hi all,
> > 
> > This is the test report of xen-unstable tree. 
> > There is 2 issues found and no issue got fixed.
> > 
> > Version Info:
> > =================================================================
> > xen-changeset:   25605:9950f2dc2ee6
> > Dom0:          linux.git  3.4.4
> > =================================================================
> > 
> > New issues(2)
> > ==============
> [...]
> > 2. Dom0 cannot be shutdown before PCI detachment from guest and when pci assignment conflicts
> >   http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1826
> 
> I think there are two issues here. The first is that xl even allows you
> to assign the same device twice (I have included this as a BUG on this
> weeks TODO posting) and the second sounds like a kernel side / pciback
> bug). Konrad does that sound plausible?

The pciback has logic to call this:

1666 /* Check whether the IRQ line is shared with other guests. */
1667 int xen_test_irq_shared(int irq)
1668 {
1669         struct irq_info *info = info_for_irq(irq);
1670         struct physdev_irq_status_query irq_status = { .irq = info->u.pirq.pirq };
1671 
1672         if (HYPERVISOR_physdev_op(PHYSDEVOP_irq_status_query, &irq_status))
1673                 return 0;
1674         return !(irq_status.flags & XENIRQSTAT_shared);
1675 }
1676 EXPORT_SYMBOL_GPL(xen_test_irq_shared);


to figure out whether it needs to Ack the interrupt line in dom0 - to avoid
those 'nobody cared' errors.

Could it be that the XENIRQSTAT_shared is not being stampted anymore?

> 
> Ian.
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel